基于双边博弈的多AUV协同对抗机理与方法研究

This project studies the cooperative confrontation problem for autonomous underwater vehicles (AUVs). Different from the cooperative searching problem, the cooperative confrontation issue more appears as the game features between two confrontation sides. However, the mechanisms and methods for multiple AUV confrontation remain unclear at present. As the game relies on the means of destruction and benefit trade-offs, sonar constrains becomes the key restriction for game confrontation due to the limitation of underwater environment. Information constraint causes the nonhonolomic constraint of game strategy as well as the ambiguity of game benefits. These factors influence the game confrontation in a very complex way and are also very difficult to study. In this project, particle swarm optimization searching method and multi-objective convergence theory are adopted to study the cooperative confrontation problem under sonar constraint, which includes (1) game model and parameter action principles of multiple AUV cooperative confrontation; (2) Nash equilibrium algorithm for multiple AUV cooperative confrontation based on sonar constraint; (3) multiple AUV game confrontation under uncertain benefit information; (4) experimental validation of cooperative confrontation for multiple AUVs. The aim of this project is to reveal the game mechanisms of multiple AUV confrontation, obtain the benefit principle for multiple AUV confrontation, master the Nash equilibrium solving algorithm of confrontation game, elucidate the sonar constraint influence on Nash equilibrium and equilibrium solution, and determine the corresponding boundary conditions, which is expected to provide theoretical foundation for engineering practice.

本项目研究多AUV协同对抗问题，与协同搜索不同，协同对抗更多地表现出编队双方之间的“博弈”性，目前对多AUV博弈的机理与方法等尚不清楚。博弈依赖于毁伤手段和收益权衡，由于水下环境特点，声呐约束对AUV对抗博弈具有关键作用，造成博弈策略非完整的多源约束以及博弈收益的不明确等科学问题，这些因素对AUV博弈的影响非常复杂，具有很大的研究难度。本项目主要采用粒子群优化搜索方法与多目标收敛理论，研究声呐约束下AUV编队之间的对抗博弈。包括：(1)多AUV协同对抗的博弈模型与参数作用机理；(2)基于声呐约束的多AUV协同对抗博弈纳什均衡算法；(3)基于收益不确定的多AUV对抗博弈；(4)协同对抗博弈实验验证等。以揭示多AUV协同对抗的博弈机理；获取多AUV协同对抗博弈的收益规律；掌握对抗博弈的纳什均衡求解算法；明确声呐约束对博弈均衡和均衡求解的影响，并确立相应的边界条件，为水下无人对抗提供理论依据。

本课题以多AUV协同对抗为研究对象，从毁伤与非毁伤两种博弈对抗角度出发，依据不同博弈策略建立了多种博弈对抗模型。以多AUV的多次对抗为作战背景，同时考虑敌我双方对抗策略，提出多AUV的动态协同攻防对抗方法；针对多AUV的协同对抗博弈受声呐性能显著约束这一问题，对声呐信号特点分析，建立声呐约束下的不确定博弈对抗模型；从非毁伤角度出发，针对单一目标提出了入侵者指向和动态位置指向的两种同步拦截博弈对抗方法；针对多个目标提出了一种基于径向基隐函数图案生成的动态多目标自适应围捕博弈算法。. 针对多AUV协同对抗的特点，综合考虑了毁伤对抗与非毁伤对抗两个方面，结合多战斗回合与声呐约束，对方目标的数量，拦截成功的上界条件，静态环境、分裂-聚集环境及障碍物环境等，系统深入地论证了多AUV协同对抗博弈方法。. 首先，研究了多AUV毁伤对抗博弈问题，以多AUV的多次对抗为作战背景，从同时考虑敌我双方对抗策略的角度出发，考虑生存概率指标函数和水下环境影响，建立了基于动态博弈的多自主水下航行器的单元目标分配模型，采用粒子群算法求解该博弈的纳什均衡，形成博弈对抗双方的最优攻防决策方案。.其次，针对声呐探测能力约束带来的博弈收益不确定问题，基于声呐信号的特点，首先从探测系统发射端考虑，结合跳频传输方式，改进信号的探测距离及速度测量结果，然后从信号处理算法考虑，采用信号恢复与分离的处理算法，引入不确定性影响因子，建立声呐约束下的不确定博弈对抗模型。. 然后，依据对方目标的数量，从非毁伤角度出发，针对单一目标提出入侵者指向和动态位置指向的两种同步拦截博弈对抗方法，并分析拦截成功的上界条件，针对多个目标提出一种基于径向基隐函数图案描述方法，针对不同环境可自适应生成连续变化的围捕图案，实现对动态多目标的围捕任务。. 最后，利用E-puck和NPU机器人搭建博弈避险试验平台并进行实验，实验结果验证了群集机器人博弈的可实施性。